Lithos 109 (2009) 72–80

Contents lists available at ScienceDirect

Lithos

j ourna l homepage: www.e lsev ie r.com/ locate / l i thos

Kimberlites and aillikites as probes of the continental lithospheric mantle

Don Francis ⁎, Michael PattersonEarth and Planetary Sciences, McGill University, GEOTOP UQAM-McGill, Montréal, Québec, Canada H3A 2A7

⁎ Corresponding author.E-mail address: donf@eps.mcgill.ca (D. Francis).

0024-4937/$ – see front matter © 2008 Elsevier B.V. Aldoi:10.1016/j.lithos.2008.05.007

A B S T R A C T

A R T I C L E I N F O

Article history:

Although the mantle xenoli Received 2 April 2008Accepted 15 May 2008Available online 4 June 2008

Keywords:KimberliteAillikiteLithospheric mantleContinentDiamondCarbonatite

ths carried by kimberlites are the source of much of our information about thecomposition of the mantle beneath the continents, the compositions of kimberlites themselves have receivedlittle attention for the information they carry about the nature of the lithospheric mantle. This neglect in partreflects their common fragmental, contaminated, and hybrid nature, but also the pervasive view that Group-Ikimberlites are sourced in the underlying asthenosphere. Insight into the nature of kimberlites and theirrelationship to the other alkaline ultramafic rocks, such as aillikites, olivine lamproites, and meimechites, canbe obtained by comparing their major element compositions in a way that treats their carbonate content as aprimary magmatic phase. Group-I kimberlites and aillikites contain significant magmatic carbonate and theircompositions fall to the Si-poor side of the composition of olivine. Group-I kimberlite can be distinguishedfrom aillikite on the basis of Fe content, but there appears to be a gradation between these two end-members. In contrast, olivine lamproites and meimechites contain relatively little primary magmaticcarbonate and have compositions that are more Si-rich than olivine.Pearce element ratio analysis assuming P as a conserved element indicates that much of the major elementvariation in hypabyssal kimberlites can be explained by variable amounts of olivine and orthopyroxene inproportions (∼70/30) similar to that of cratonic mantle xenoliths. Much of the olivine is present asxenocrysts, but the orthopyroxene is occult and has presumably been assimilated. The fact that individualfields of alkaline ultramafic rocks are characterized by uniform Fe and Ti contents that can be mapped on aregional scale suggests that the major element composition of these unusual rocks, and Group-I kimberlitesin particular, is a reflection of the continental lithospheric mantle with which they have interacted. Theassociation of Fe-rich aillikitic magmas with zones of cratonic rifting, and the requirements of Fe–Mgpartitioning indicate that they form deeper than Group-I kimberlites, near the intersection of the 1350 °Cmantle adiabat with the CO2 mantle solidus, at pressures of 8+GPa. According to our working model, Group-Ikimberlites are produced by the influx and reaction of carbonate-rich magmas with the highly magnesianharzburgites of the lithospheric mantle beneath continental cratons. In non-cratonic environments, theserising carbonate-rich magmas evolve into aillikites because of the lower Mg# of the asthenospheric mantle.They rarely reach the surface, however, but become the enriched component incorporated into higher-degreebasaltic melts. An inverse correlation between diamond grade and kimberlite Fe and Ti contents may simplyreflect the fact kimberlites with the lowest Ti contents have interacted with the most depleted and reducedharzburgites of the lithospheric mantle, where diamonds are most likely to be encountered.

© 2008 Elsevier B.V. All rights reserved.

1. Introduction

Despite their importance as the carriers of xenoliths of thecontinental lithospheric mantle, the compositions of kimberlitesthemselves have been relatively neglected for the information theycarry about the nature of the lithospheric mantle. This lacuna reflectsnot only the frequent fragmental, contaminated, and hybrid nature ofkimberlites, but also a commonplace held by both industry andacademia that kimberlites are sourced in the underlying astheno-sphere, and act simply as elevators that bring accidental fragments of

l rights reserved.

the mantle and diamonds to the surface. With few exceptions(Malkovets et al., 2007), conventional wisdom interprets diamondsto be xenoliths derived from the lithospheric mantle that areunrelated to their host kimberlites, which are sourced in theasthenosphere. To complicate matters, kimberlites are part of a largergroup of alkaline ultramafic rocks that include ultramafic lampro-phyres (variously termed aillikites, melnoites, etc., (Rock, 1986)), aswell as olivine lamproites, andmeimechites (Arndt et al., 1995), whosedistinction requires a detailed knowledge of their mineral paragenesisand mineral chemistry.

Until recently, a relatively small dataset has limited our knowledgeof the major element compositions of the magmas responsible for thealkaline ultramafic rocks. In this paper, we use a survey of recent

Table 1The four suites of alkaline ultramafic rocks

Definitions taken from: (Rock, 1986; Mitchell and Bergman, 1991a; Arndt et al.,1995; Mitchell, 1995; Tappe et al., 2005).

Fig. 1. a) Photomicrograph of magmatic calcite laths in the matrix of a hypabyssal-faciesNikos Kimberlite, Somerset Island. Field of view=4 mm. b) Photomicrograph ofmagmatic calcite oikocrysts in the matrix of a hypabyssal-facies Renard Kimberlite,Otish Mountains. Field of view=4 mm.

73D. Francis, M. Patterson / Lithos 109 (2009) 72–80

chemical analyses of hypabyssal-facies alkaline ultramafic rocks toshow that Group-I kimberlites and aillikites may be best understoodas mixtures of carbonate and peridotite of variableMg# (Mg/(Mg+Fe)).We demonstrate that the hypabyssal facies of the diverse members ofthe alkaline ultramafic rocks can in most cases be reliably distinguishedon the basis of their Si and Fe contents in a whole-rock major elementanalysis, if reasonable care is taken to exclude crustal xenoliths, and iftheir CO2 content is assumed to represent primarymagmatic carbonate.We go on to show that the geographic distribution of Fe and Ti contentsin alkaline ultramafic rocks appears to map regional lithosphericdomains in the mantle beneath North America. Furthermore, weconfirm the existence of an anti-correlation between the diamondproduction of kimberlite fields and the Fe and Ti contents in their hostkimberlite and conclude that, despite their sub-lithospheric origin, themajor element composition of kimberlite is established by the litho-spheric mantle they have consumed rather than the asthenosphere andthus carries important information about both their diamond potentialand the continental lithospheric mantle.

2. Alkaline ultramafic rocks

2.1. Background

Kimberlites are members of a larger group of alkaline ultramaficrocks which, despite their diversity, share the unusual compositionalcharacteristic of being highly enriched in incompatible trace elements,but highly refractory in terms of bothmajor elements, such as Mg, andcompatible trace elements, such as Cr and Ni. The nomenclature,classification, and thus petrogenesis, of these diverse alkalineultramafic rocks are a confusing legacy of a time when modernanalytical techniques were not available and formal classification wasbased upon mineralogy, and even geographic parochialism (Rock,1986). Table 1 presents a simplification of the ideas presented byothers (Rock, 1986; Foley, 1992; Mitchell, 1995) in which the alkalineultramafic rocks are grouped into 4 distinct magmatic suites, namedhere after their most primitive (highest MgO) member, with noattempt to account for their diverse mineral-based memberships, northeir more evolved (less magnesian) associates. These 4 alkalineultramafic magmatic suites are:

1. Group-I kimberlites typically corresponding to the “basaltic”kimberlites of South Africa, but by formal definition characterizedby ∼ Bulk Earth Sr and Nd isotopic ratios (Smith, 1983; Mitchell,1995; Mitchell, 2008).

2. Olivine Lamproites (Mitchell and Bergman,1991a) and “micaceous”kimberlites.

3. Ultramafic lamprophyres (Rock, 1986; Tappe et al., 2005) whosemost magnesian members are aillikites or melnoites.

4. Meimechites (Arndt et al., 1995), which grade to alkaline picrites(Arndt et al., 1998; Bernstein et al., 2000).

South African kimberlites have historically been divided intobasaltic and micaceous sub-types. This subdivision was reworked bySmith (Smith, 1983) who subdivided kimberlites into Group-Ikimberlites whose Sr and Nd isotopic values are similar to those ofBulk Earth and Group-II kimberlites with more radiogenic strontiumand less radiogenic Nd isotopic ratios, similar to those of EM2. Ingeneral, Group-I kimberlites are equivalent to basaltic kimberlites, butthe existing data for South African Group-II kimberlites appear toinclude both micaceous and basaltic kimberlites, and for this reasonthey have not been included in this review. Furthermore, kimberliteswith transitional isotopic compositions are not uncommon (Skinner,et al., 1994) and there are a number of inconsistencies, for example;the Aries Pipe of Australia has isotopic characteristics similar to SouthAfrican Group-I kimberlites, but is compositionally and mineralogi-cally more similar to an olivine lamproite or micaceous kimberlite(Edwards et al., 1992). There are also numerous disagreements; forexample the kimberlites of western Greenland (Larsen and Rex, 1992)are ultramafic lamprophyres according to Mitchell (Mitchell, 1986).The situation is complicated by the problems of heteromorphism

Fig. 2. K versus Si in cation units. Symbols: Group-I kimberlites — open circle, olivinelamproite — shaded diamond, aillikite — black triangle, meimechite — shaded square.Data Sources: (Jaques et al., 1984; Bourne and Bossé, 1991; Edwards et al., 1992; Fraserand Hawkesworth, 1992; Larsen and Rex, 1992; Taylor et al., 1994; Graham et al., 1999;Schmidberger and Francis, 1999; Digonnet et al., 2000; Price et al., 2000; Baragar, Maderand Lecheminant, 2001; Kaminsky et al., 2002; Le Roex et al., 2003; Lewis et al., 2003;Birkett et al., 2004; Mitchell, 2004; Nowicki et al., 2004; Tappe et al., 2004; Becker andRoex, 2006; Tappe et al., 2006).

Fig. 3. Si versus C in cation units for Group-I kimberlites and aillikites. Symbols: naturalcarbonatites — open inverted triangle, experimental melts of Fe-free carbonatedperidotite — shaded St. Andrew's cross (Dalton and Presnall, 1998; Moore and Wood,1998), other symbols as in Fig. 2. The degree of melting (0–1%) of the experimental meltsincreases with increasing Si.

74 D. Francis, M. Patterson / Lithos 109 (2009) 72–80

associated with the volatile-dependent stabilities of phases such ascarbonate and phlogopite, which leads to rocks with otherwiseidentical volatile-free compositions having different names (Yoder,1986). The common presence of incomplete reactions involvinghydrous or carbonated phases exacerbates the problem. The loss ofCO2 in kimberlites leads to the formation of monticellite and diopsiderather than calcite, and the loss of water in olivine lamproites leads tothe loss of phlogopite and the appearance of pyroxenes. Theseproblems have lead to a plethora of rock names in classificationschemes based on mineralogy (Rock, 1986; Yoder, 1986). To makematters worse, the common fragmental nature of these rocks,especially where they occur as diatreme breccias with clastic matricesor crater-facies tuffaceous volcano-sediments, along with the abun-dance of accidental fragments of both crustal and mantle lithologies,as well as their susceptibility toweathering and serpentinization, havediscouraged studies of the whole rock chemistry of kimberlites. As aresult, the classification of any given occurrence remains a ratheresoteric process, often best left to experts.

One of the most important features of Group-I kimberlites andaillikites is the presence of abundant carbonate, whereas carbonateappears to be minor or secondary in olivine lamproites (Mitchell andBergman, 1991b) and meimechites (Arndt et al., 1995). Carbonatefrequently occurs as late-stage segregations and/or secondary repla-cement in altered kimberlite and aillikite; but in pristine hypabyssalsamples carbonate minerals commonly appear to be magmatic,exhibiting crystal habits ranging from euhedral microphenocrysts(Fig. 1a) in a fine-grained matrix to large oikocrystic crystals (Fig. 1b)enclosing rounded olivine. A number of studies have argued theprimarymagmatic nature of such kimberlite carbonate using textures,trace elements, and oxygen and carbon isotopes (Kirkley et al., 1989;Armstrong et al., 2004; Kamenetsky et al., 2007; Wilson et al., 2007).The carbonate in Group-I kimberlites is dominantly calcite, with lesserdolomite, with the most CO2-rich and SiO2-poor examples beingtermed calico-kimberlites (Mitchell, 1995). In contrast, the carbonateof aillikites is more commonly dolomite to ankerite (Rock, 1986). Thismagmatic carbonate must be taken into account when interpretingthe chemical variation diagrams of kimberlitic rocks, but unfortu-

nately many consider carbonate mineralization to be secondary andtoo frequently published analyses of kimberlites do not report CO2.

2.2. The compositions of alkaline ultramafic rocks

The recent appearance of a number of studies of the fine-grainedhypabyssal facies of kimberlites and aillikites whose analyses includeCO2 has made it possible to examine their compositional systematicsfor evidence of the involvement of carbonate phases. The analyses inthis assembled dataset (see Fig. 2 for data sources) are classifiedaccording to their original sources, with the analyses of Group-Ikimberlites being filtered using the crustal contamination index((SiO2+Al2O3+Na2O)/ (2×K2O+MgO)) (Clement, et al., 1984; Skinneret al., 1994). These analyses have been recalculated into cation units,including carbon calculated from their CO2 contents, and thennormalized to a sum of 100 cations. In principle, one should alsoinclude the H of the magmatic water in kimberlite in the cationcalculation. The amount of magmatic water in kimberlite is, however,hard to accurately constrain. The reported water contents of theGroup-I kimberlites in our data set ranges from 1 to 19 wt.%, with anaverage of ∼8 wt.%. Hydrogen and oxygen isotope studies ofgroundmass serpentine in kimberlites suggests, however, that muchof this water is meteoric (Sheppard and Dawson, 1975), as doesserpentine mineral paragenesis (Stripp, et al., 2006) which suggeststhat serpentinization involves post-magmatic fluids that are poor inCO2. The magmatic water content of kimberlites is thus likely to beconsiderably lower than the 5 to 10 wt.% commonly reported inanalyses. In any case, ignoring water will only cause problems if thereare hydrous phases at the magmatic stage. The composition ofphlogopite, for example, changes from being slightly higher in Si (37.5cations) than olivine (33.33), to slightly lower (30.0), if H is included inthe cation total, but the position of the phlogopite-olivine join is littlechanged. The problem would be much more severe if serpentine(Si=22.22) were a magmatic phase, but this seems unlikely in light ofserpentine's upper thermal stability limit (∼500 °C) (Stripp et al.,2006). For the forgoing reasons, the reported H2O contents in our dataset were not used to include H in the cation calculation.

One of the most striking features of the dataset is the presence of adistinct population minimum at ∼33 cation % Si that coincidesapproximately with the olivine–phlogopite compositional join (Fig. 2).

Fig. 5. Si versus Fe in cation units. The vertical solid line represents the composition ofolivine (Si=33.33 cation units). The dashed line on the left separates most aillikites fromGroup-I kimberlites. It starts at an olivine composition of Fo85 and extends towards thecomposition of ideal calcite (Si=0.00 cation units). The dashed line on the right starts atan olivine composition of Fo90 and separates meimechites from olivine lamproites.Symbols as in Fig. 2.

75D. Francis, M. Patterson / Lithos 109 (2009) 72–80

Both aillikites and group-I kimberlites have low Si contents that plot tothe Si-poor side of the olivine — phlogopite join (Fig. 2). This meansthat they lie outside the chemical space that can be expressed in termsof the compositions of mantle silicates alone, requiring the presence ofa Si-poor phase such as carbonate. The inclusion of water in the cationcalculation would move the aillikite and Group-I kimberlite analysesto ever lower Si cation contents. In a plot of C versus Si, the analyses ofGroup-I kimberlites and aillikites plot along a line between carbonateand olivine (Fig. 3). In contrast, the olivine lamproites and meime-chites (Fig. 2) are more silica rich than olivine, such that theircompositions can be expressed in terms of mantle silicates alone, withno requirement for carbonate.

3. Discussion

3.1. Implications for the identification and classification of the alkalineultramafic rocks

Each of the four ultramafic alkaline suites defines a large range inMg content with relatively less variation in Fe (Fig. 4). In Fe–Mgchemical space, however, both olivine lamproites and all Group-IIkimberlites plotwith Group-I kimberlites at low Fe contents, while theaillikites and meimechites now plot together at distinctly higher Fecontents. This reversal in association means that the 4 alkalineultramafic rock suites can be effectively distinguished on the basis ofSi and Fe (Fig. 5). Although it is apparent that there is a continuousgradation in Fe content on the Si-poor side of the olivine-phlogopitejoin from the low values in classical Group-I kimberlite suites(FeOb11 wt.%) to Fe-rich kimberlites and aillikites (FeON12 wt.%),with little regard for Mg (Fig. 4). A line joining the composition ofcarbonate to olivine with a composition of Fo85 effectively separatesall aillikites from themajority of Group-I kimberlites (Fig. 5). The high-Fe Group-I kimberlites that do plot with aillikites in terms of Fe, alsoshare their relatively higher K (Fig. 2) and Al contents, and arecompositionally indistinguishable from them. A similar range in Fecontent exists on the Si-rich side of the olivine-phlogopite join, with

Fig. 4. Fe versus Mg in cation units. Symbols as in Fig. 2, except that mantle xenolithsfrom kimberlites and Cordilleran alkaline basalts are represented by black crosses(Francis, 2003). The solid line represents the composition of olivine (Fe+Mg=66.67cation units). The dashed lines indicate the compositions of liquids that wouldequilibrate with the indicated olivine Fo contents: 0.95, 0.90, and 0.85, assuming Fe/Mgoliv / Fe/Mgliq=0.33 (Roeder and Emslie, 1970). The dotted lines indicate the slopes ofisotherms, with the range of temperature between the upper and lower dotted linesbeing ∼200 °C at any given pressure (the actual temperature depends on the pressure).P is the composition of Primitive Mantle (O'Neill and Palme, 1998; Palme and O'Neil,2003).

olivine lamproites constituting an array at low Fe contents that iseffectively separated from more Fe-rich meimechites by a lineoriginating at an olivine composition of Fo90 (Fig. 5).

The coincidence of a population minimum in the alkalineultramafic rock dataset with the olivine–phlogopite join, combinedwith the common presence of these phases as macrocrysts in theserocks, suggest that this join represents an alkemade line or thermaldivide that provides a theoretical justification for distinguishingbetween “basaltic” Group-I kimberlites and olivine lamproites (Figs. 2and 5) on the basis of their Si content. Group-I kimberlites have less Sithan olivine, while olivine lamproites have more Si than olivine. Thecomposition of olivine divides the alkaline ultramafic rocks into twofundamentally different potential liquid lines of descent, one towardsincreasing Si, the other towards decreasing Si. Similarly, the Si contentof olivine provides a fundamental criterion for distinguishing betweenaillikites and meimechites. The origin of the population minimum inthe data set could reflect the tendency of residual liquids on either sideof the join to move away from each other by the fractionation ofolivine plus or minus phlogopite on route to the surface. Fractionationof the olivine and phlogopite phenocrysts that characterize many ofthese rocks would lead to more Si-poor, but carbonate-rich residualliquids from Group-I kimberlite and aillikite parental magmas, and Si-rich residual liquids from olivine lamproites and meimechites, whichis consistent with the common association of olivine lamproites withmore Si-rich lamproite derivatives such as leucite lamproites (Jaqueset al., 1984). Despite the foregoing arguments, however, the variationof Si in Group-I kimberlites and aillikites cannot be produced byolivine fractionation because Al remains constant or decreases withdecreasing Mg and Si, as C increases. Pearce element ratio analysisusing P (Fig. 6) or Ti as the conserved element (Russell and Nicholls,1988) indicates that much of the major element variation inhypabyssal kimberlites can be explained by adding variable amountsof olivine and orthopyroxene in proportions (∼70/30) similar to thoseof harzburgite xenoliths to a carbonate-rich (∼20 wt.%CO2) and silica-poor (∼20 wt.% SiO2) liquid (Fig. 6). Much of this olivine likely remainsas the ubiquitous olivine xenocrysts seen in kimberlites, suggestingthat the carbonate-rich magma is saturated in olivine. Orthopyroxene,however, is rarely observed as macrocrysts in kimberlite (Mitchell,2008) and must be undersaturated in kimberlite magma. The

Fig. 7. a) Si versus Fe in cation units for the aillikitic fields of Lac Leclair (white symbols)and Demaraisville (grey symbols), Québec, Canada. b) Si versus Fe in cation units for thekimberlites of the central Slave Province, Northwest Territories, Canada. Those that fallto the Si-poor side of the olivine line are excluded from the main data base because oftheir high contamination indices.

Fig. 6. Si/P versus (Mg+Fe)/P in cation units for Group-I kimberlites. The solid straightlines are those for the stoichiometric addition or subtraction of olivine, orthopyroxene,and clinopyroxene assuming P is a conserved element. Thin dashed line— linear best fitto Group-I kimberlites (R2=0.97, slope=1.73.

76 D. Francis, M. Patterson / Lithos 109 (2009) 72–80

xenocrystic orthopyroxene must therefore have been assimilated,with a concomitant rise in the silica content of the magma.

The olivine–phlogopite join may separate two distinct primarymagma compositions, one on the Si-poor side produced at aninvariant point involving carbonate and the other produced on theSi-rich side at an invariant point involving phlogopite. Meltingexperiments on carbonated Fe-free model mantle peridotite (Daltonand Presnall, 1998; Moore and Wood, 1998) support the firstpossibility, with the continuum of haplo-kimberlite liquids producedwith increasing degree of partial melting coinciding (Fig. 3) with thecompositional range of Group-I kimberlites. At the degree of partialmelting at which the compositions of the experimental liquidsapproximate those of Group-I kimberlites (∼1 wt.%), however,carbonate had been exhausted in a peridotite source containing0.1 wt.% CO2 (Gudfinnsson and Presnall, 2005).

3.2. Compositional variations within individual fields

An examination of individual fields of kimberlite or aillikiteoccurrences indicates that although the alkaline ultramafic rocks ofany given field are characterized by relatively constant Fe-contents, Siis typically quite variable. It is not unusual for individual kimberlite oraillikite fields to contain samples on both sides of the olivine–phlogopite join (Fig. 7). It may be that the presence of both Si-poor andSi-rich alkaline ultramafic rocks in some fields reflects the coexistenceof two primary magmas, one produced at an invariant point involvingcarbonate, and the other phlogopite, in a variably carbonated andhydrated lithospheric mantle source. In many cases, however, thewide variation in Si can be attributed to secondary processes. Forexample, in aillikitic fields, such as the Leclair (Baragar et al., 2001),Demaraisville (Bourne and Bossé, 1991) (Fig. 7a) and Aillikite Bay(Tappe et al., 2006), the coexistence of carbonatite, aillikite, meime-chite, indicates that the Si variation within the suites may in partinvolve the variable loss of CO2 as an immiscible carbonate fluid(Tappe et al., 2006). In contrast, the Si-rich kimberlites of the centralSlave Province (Fig. 7b) have high contamination indices (Clementet al., 1984; Skinner et al., 1994) indicating significant interaction withthe crust.

Individual fields of dykes and/or pipes of alkaline ultramafic rocksare characterized by relatively uniform Fe and Ti contents. Theexistence of highly magnesian aillikites defeats attempts to distin-

guish them fromGroup-I kimberlites on the basis of Ca/Mg ratio (Rock,1986), but aillikites and meimechites are always richer in Fe and Tithan Group-I kimberlites and olivine lamproites, and a Fe limit of∼11 wt.% FeO reliably separates most aillikites. This distinction issomewhat arbitrary, however, because a number of Kaapvaal cratonGroup-I kimberlites would be classified as aillikites according to theirFe content (Figs. 5 and 9) by our scheme, and there is disagreementabout whether magnesian aillikites are, or are not, Fe-rich Group-Ikimberlites. The important point is that, although there is acontinuous range in Fe and Ti contents from the low values forclassical Kimberly basaltic kimberlites to Fe-rich aillikites, individualfields are characterized by relatively restricted ranges in Fe, despiteexhibiting relatively large variations in Si.

3.3. Regional distribution of kimberlite, aillikite, and olivine lamproite

Unlike Si, the relative constancy of Fe and Ti in any given field ofalkaline ultramafic dykes and/or pipes suggests that their Fe-contentreflects the Fe-content of their mantle source. Insight into the physicalrelationship between the Fe-poor mantle sources presumed forkimberlites versus the more Fe-rich mantle sources presumed for

Fig. 8. Alkaline ultramafic suite locations plotted on a radar image of Canada showing their position with respect to major tectonic boundaries. Group-I kimberlites — white circle,olivine lamproites — white diamond, aillikite/meimechite — black triangle.

77D. Francis, M. Patterson / Lithos 109 (2009) 72–80

aillikites and/or meimechites can be obtained from the regionalgeographical distribution of these different alkaline ultramafic rocks.A spatial association between olivine lamproites and kimberlites andcratonic mantle roots has been demonstrated in Australia (Jaques andMilligan, 2004). A survey of the known occurrences of alkalineultramafic rocks on the North American continent (Fig. 8) clearlysuggests the existence of regional compositional domains in mantlesource regions. The most striking is the preferential association ofaillikites with the rift between Greenland and Labrador, extendinginto the central Ungava Peninsula, along the Grenville Front, and alongthe northeast shore of Lac Superior associated with the MidcontinentRift (Queen et al., 1996). In contrast, the interiors of the Slave andSouthern Superior cratons are dominated by Fe-poor Group-Ikimberlites, whereas olivine lamproites and related rocks arepreferentially distributed along the eastern edge of the NorthAmerican Cordillera and western Greenland. Similarly, the Group-Ikimberlites of the Kaapvaal craton of South Africa actually appear torange in composition from the type Group-I kimberlites of Kimberley

Fig. 9. Si versus Fe in cation units of Group-I kimberlites of the Kaapvaal craton. In thisdiagram, Kaapvaal kimberlites that fall in the aillikite field are shown as black triangles.Other symbols as in Fig. 2.

(Le Roex et al., 2003) to more Fe-rich aillikitic compositions, such asthose of the Gansfontein field (Becker and Roex, 2006) (Fig. 9).

3.4. Implications for the mantle sources of kimberlite and aillikite

The existence of regional domains (Fig. 8) in the Fe and Ti contentsof alkaline ultramafic rocks poses a problem for models that source

Fig. 10. Pressure — temperature diagram showing three Canadian mantle xenolithgeothermswith themineral andmelting equilibria for mantle peridotite under differentvolatile contents and oxidation states, all superimposed on the region (grey field)between the top of the low velocity zone and the Lehmann discontinuity. Squares —

Diavik mantle xenoliths, central Slave Province (Graham et al., 1999), circles— SomersetIsland Xenoliths from Nunavut (Schmidberger and Francis, 1999), and diamonds —

Cordilleran mantle xenoliths from alkaline basalts along the Canadian Cordillera (Shiet al., 1998). The xenolith equilibration temperatures were calculated with theclinopyroxene geothermometer of Nimis and Taylor (Nimis and Taylor, 2000), and thepressures were calculated using the Finnerty and Boyd calibration (Finnerty and Boyd,1987) of the MacGregor Al in orthopyroxene geobarometer (MacGregor, 1974). Sourcesfor the labeled reactions: dry solidus (Hirschmann, 2000), CO2 solidus (Dasgupta andHirschmann, 2007), wet solidus (Katz et al., 2003), CH4 solidus (Taylor and Green,1988).Cross-hatched grey region indicates the stability field of amphibole, the stability field ofcarbonate at oxygen fugacities greater than 2 log units below the FMQ buffer is indicatedby the region with inclined grey-lines.

Fig. 11. Mg/Si versus Al/Si in cation units for Group-I kimberlites (open circles) andaillikites (black triangles). Other symbols: P — primitive mantle, shaded diamonds —

primitive MORB, kimberlite low temperature mantle xenoliths and Cordilleran alkalinebasalt xenoliths — black crosses, carbonaceous chondrites — dotted squares, ordinarychondrites — black squares, crossed squares — enstatite chondites (Francis, 2003).

Fig. 12. A working model for the genetic relationship between Group-I kimberlites andaillikites.

78 D. Francis, M. Patterson / Lithos 109 (2009) 72–80

Group-I kimberlites in the sub-lithospheric mantle. Appeals todifferent depths of melting to explain the Fe differences betweenkimberlites and aillikites leads to further problems. If Fe-rich aillikiticmagmas are associated with rifts, then at first blush one would thinkthey would be more shallowly sourced than Group-I kimberlitessourced in the asthenosphere beneath thick continental lithospheres.However, because of the nature of Fe–Mg partitioning between olivineand liquid (Fig. 4), in order for the Fe content of aillikites to beproduced by melting a mantle source with the same Mg# as the onethat produced the Group-I kimberlites, the melting for the aillikiteswould have to occur at significantly higher temperatures, and thussignificantly deeper if melting is adiabatic (Fig. 10). If this were ageneral scenario, one would expect aillikites to be more magnesianthan Group-I kimberlites, rather than the reverse. No amount ofjuggling temperature and pressure will generate the different Fe andTi contents of Group-I kimberlites and aillikites from the same mantlesource, unless the aillikites formmuch deeper. Themore Fe and Ti-richaillikites and meimechites, however, are too Fe-rich to haveequilibrated with primitive mantle (Mg# ∼0.88) at lithosphericdepths, much less the more magnesian composition (Mg# ∼0.93) ofrefractory cratonic roots.

Further insight into the source regions of kimberlites can beobtained by a comparison of the calculated equilibration pressure andtemperatures of mantle xenoliths with the most recent determina-tions of the mantle solidus under various oxidation states and volatilecompositions (Fig. 10). There is a striking correspondence between thetemperature estimates for the deepest mantle xenoliths and thetemperature region between the CH4 and CO2 experimental solidii(Fig. 10). The deepest mantle xenoliths in kimberlites record very lowoxidation states (∼ IW) (McCammon and Kopylova, 2004; Woodlandand Koch, 2003), at which carbon will be in the form of diamond and/or methane (CH4). Under such conditions, the mantle would melt atthe CH4 solidus (Taylor and Green, 1988), which is crossed by the1350 °C mantle adiabat at pressures of approximately 7 GPa. CO2

introduced into such a reduced refractory mantle would initially bereduced to form diamond (Malkovets et al., 2007), but as the influxcontinued, the oxidation state of the peridotite would rise untilcarbonate is stabilized. The formation of carbonate may haveimportant implications for continental free board because of theassociated decrease in density of the refractory mantle root.Furthermore, this carbonated mantle would undergo partial meltingat the much lower temperature of the CO2 solidus, and the absence of

mantle xenoliths with pressure-temperature estimates above the CO2

solidus suggests that this may be the general case in themantle sourceregions of Group-I kimberlites.

The preferential association of Fe and Ti rich aillikites with zones ofrifting, weakness, and/or the peripheries of cratons is more enigmatic.The compositions of aillikites and meimechites are too Fe-rich to haveequilibrated with primitive mantle at lithospheric depths, a fact thatcould be used to argue for an Fe-rich component in the mantlebeneath West Greenland and Labrador, and perhaps peripheral tocontinental cratons in general, such as the perisphere of Anderson(Anderson, 1995). Refractory spinel harzburgites xenoliths in Tertiaryalkaline basalts of both East (Bernstein et al., 1998) and West(Bernstein and Brooks, 1998) Greenland, however, are characterizedby very magnesian olivines (Fo90–94), similar to those of cratonicmantle xenoliths in kimberlites. Furthermore, 600 Ma aillikites (or Fe-rich kimberlites) in West Greenland carry refractory garnet and spinelharzburgite to lherzolite xenoliths with Fo90–94 olivine (Bizzarro andStevenson, 2003). Some aillikite dykes do contain more Fe-rich duniteand olivine clinopyroxene xenoliths, but these are interpreted to becumulates (Bernstein and Brooks, 1998; Bizzarro and Stevenson,2003). There is thus little evidence of anomalous Fe-rich mantlebeneath Greenland and Labrador. An alternate explanation for the Fe-rich character of aillikite and meimechite magmas might be muchgreater depths of origin, approaching the intersection of the mantle1350 °C adiabat with the CO2 solidus. The solidus temperatures atthese depths are so high (∼1400 °C) that the initial melts of a mantlewith a Mg# of 0.88 will have Fe contents approaching those ofaillikites. Such a possibility is supported by the fact that thecompositions of aillikites coincides with that of primitive mantle ina plot of Mg/Si versus Al/Si, while Group-I kimberlites are displacedtowards the more magnesian compositions (Fig. 11) of refractorycratonic mantle xenoliths.

Fig. 13. Diamond grade in carats per hundred ton for a pipe versus the wt.% TiO2 of thehost kimberlite. Symbols: Canadian Group-I kimberlites — open circles, Kaapvaalkimberlites — grey squares, Siberian kimberlites — Crosses (Vasilenko et al., 2002).

79D. Francis, M. Patterson / Lithos 109 (2009) 72–80

3.5. A working model for the relationship between kimberlite and aillikite

We propose the following working model to explain the origin andrelationship of Group-I kimberlites and aillikites (Fig. 12). Underoxidizing conditions, adiabatically-rising fertile mantle (Mg# ∼0.88)crosses its CO2 solidus at pressures between 8 to 10 GPa (∼300 km), andan upwardly mobile carbonate-rich fluid is released. Such carbonate-rich melts would evolve to aillikitic melts as the degree of meltingincreases with decreasing pressure. Where the carbonate-rich meltsencounter the refractory keels beneath Archean cratons, however, theyinteract with more magnesian harzburgite with a Mg# of ∼0.93, andthey evolve into kimberlites by a combination of wall rock reaction andbulk assimilation of the continental lithospheric mantle. The extent towhich this interaction involveswholesalemechanical contamination bydisaggregated lithospheric peridotite rather than chemical reaction ispoorly constrained, but to a first approximation the compositionalvariation in Group-1 kimberlites can be reproduced by mixtures acarbonate-rich magma and peridotite with an olivine/orthopyroxeneratio of ∼70/30. Much of the olivine remains as the olivine xenocrysts,however, the orthopyroxene is occult and appears to have beenassimilated. Regardless of the degree to which the compositions ofGroup-1 kimberlites represent liquids, however, their isotopic composi-tion will be inherited from the trace-element-enriched carbonate fluidsourced beneath the lithosphere, but theirmajor element characteristicsreflect those of the lithosphericmantle they have consumed. The low Fecontents of Group-I kimberlites reflect the highly refractory nature(Mg# ∼0.93) of the continental lithosphere mantle with which risingCO2-rich melts interact. Aillikitic magmas which avoid the refractorylithospheric mantle roots beneath the continents will become poorer inMg and Fe as they interact with the solid silicate mantle with droppingpressure and temperature because of their effect on olivine partitioning,and would likely become incorporated into larger degree partial melts,perhaps becoming the enriched component observed inMORB. The factthat the range in Fe content fromGroup-I kimberlites to aillikites exactlyencompasses the range of Fe contents of terrestrial picritic basalts is nota coincidence in our model, but simply reflects the range of mantle Mgnumbers. In this light, aillikites andGroup-I kimberlitesmight be viewedas the enriched component in many terrestrial basalts, in a manneranalogous to KREEP in lunar basalts (Anderson, 1989).

3.6. Implications for diamond exploration

Our conclusion that hypabyssal samples of the diverse alkalineultramafic rocks can be reliably distinguished on the basis of whole rockchemical analysis has obvious application to the evaluation of the

diamond potential of alkaline ultramafic rocks. Furthermore, ourupdated dataset supports earlier claims (Vasilenko et al., 2002) thatthere is a relationship between the bulk composition of a kimberlite andits potential diamond grade, with the best diamond grades associatedwith the lowest bulk rock Fe and Ti contents (Fig. 13) — a findingconsistent with the rule of thumb of old-timers that kimberlites rich intitano-magnetite do not carry diamonds. This anti-correlation defiesconventionalwisdom that kimberlites simply transport diamonds to thesurface, and are genetically unrelated to their precious cargo. The factthat there appears to be a continuous variation in Fe and Ti betweenGroup-I kimberlites and aillikites means that determining the Fe and Tiof a kimberlite could be an effective tool in evaluating its diamondpotential. Furthermore, the apparent existence of lithospheric domainsin Fe and Ti contents means that a systematic mapping of thecompositions of alkaline ultramafic rocks could be a regional scaleexploration tool. The association of the highest diamond grades withkimberlites with the lowest Fe and Ti contents may simply reflect thefact that such kimberlites have interacted with the most reduced anddepleted harzburgites of the lithospheric mantle, where diamonds aremost likely to be encountered.

4. Conclusions

The major element compositions of hypabyssal-facies alkalineultramafic rocks not only provide relatively simple criteria for theiridentification and subdivision, but also a means of probing thecomposition of the lithospheric mantle. The Si content of olivine is aneffective means for distinguishing between Group-I kimberlites andolivine lamproites, and between aillikites and meimechites, while Fecontent distinguishes Group-I kimberlites from aillikites and olivinelamproites from meimechites. Our results indicate that magmaticcarbonate plays an essential role in the petrogenesis of Group-Ikimberlites and aillikites. Although kimberlite magmas owe theirorigin to carbonate-rich melts rising from the asthenosphere, theirmajor element compositions reflect the lithospheric mantle they haveconsumed. Regional variations in the compositions of alkalineultramafic rocks provide information about the composition andextent of the underlying lithospheric mantle. Finally, the Fe and Ticontent of an alkaline ultramafic rock is a significant indication of itspossible diamond potential — the lower Fe and Ti the better.

Acknowledgements

The ideas developed in this manuscript have benefited greatlyfrom discussions with numerous scientists whose passion is kimber-lites. This research was supported by National Science and Engineer-ing Research Council of Canada Discovery Grant # RGPIN 7977-00.GEOTOP Publication No. 2007-0059.

References

Anderson, D., 1989. Theory of the Earth. Blackwell Scientific Publications, p. 366.Anderson, D.L., 1995. Lithosphere, asthenosphere, and perisphere. Reviews of Geophysics

33, 125–149.Armstrong, J.P., Wilson,M.R., Barnett, R.L., Nowicki, T., Kjarsgaard, B.A., 2004.Mineralogy

of primary carbonate-bearing hypabyssal kimberlite, Lac de Gras, Slave Province,Northwest Territories, Canada. Lithos 76, 415–433.

Arndt, N., Lehnert, K., Vasil'ev, Y., 1995. Meimechites: highly magnesian lithospheric-contaminated alkaline magmas from the deep subcontinental mantle. Lithos 34,41–59.

Arndt, N., Chauvel, C., Czamanske, G., Fedorenko, V., 1998. Two mantle sources, twoplumbing systems: tholeiitic and alkaline magmatism of the Maymecha RiverBasin, Siberian flood volcanic province. Contributions Mineralogy Petrology 133,297–313.

Baragar, W.R., Mader, U., Lecheminant, G.M., 2001. Paleoproterozoic carbonatitic ultrabasicvolcanic rocks (meimechites?) of Cape Smith Belt, Quebec. Canadian Journal of EarthSciences 38, 1313–1325.

Becker, M., Roex, A.P.L., 2006. Geochemistry of South African on- and off-craton, Group Iand Group II kimberlites: petrogenesis and source region evolution. Journal ofPetrology 47 (4), 673–703.

80 D. Francis, M. Patterson / Lithos 109 (2009) 72–80

Bernstein, S., Brooks, C.K., 1998. Mantle xenoliths from Tertiary lavas and dykes onUbekendt Ejland, West Greenland. Geology of Greenland Survey Bulletin 180,152–154.

Bernstein, S., Kelemen, P.B., Brooks, C.K., 1998. Depleted spinel harzburgite xenoliths intertiary dykes from East Greenland: restites from high degree melting. Earth andPlanetary Science Letters 154 (1–4), 221–235.

Bernstein, S., Leslie, A.G., Higgins, A.K., Brooks, A.K., 2000. Tertiary alkaline volcanics inthe Nunatak Region, Northeast Greenland: new observations and comparison withSiberian maymechites. Lithos 53, 1–20.

Birkett, T.C., McCandless, T.E., Hood, C.T., 2004. Petrology of the Renard igneous bodies:host rock for diamond in the northern Otish Mountains region, Québec. Lithos 76,475–490.

Bizzarro, M., Stevenson, R.K., 2003. Major element composition of the lithosphericmantle under the North Atlantic craton: evidence from peridotite xenoliths of theSarfartoq area, southwestern Greenland. Contributions Mineralogy Petrology 146,223–240.

Bourne, J.H., Bossé, J., 1991. Geochemistry of ultramafic and calc-alkaline lamprophyresfrom the Lac Shortt area, Quebec. Mineralogy and Petrology 45, 85–193.

Clement, C.R., Skinner, E.M.W., Scott Smith, B.H., 1984. Kimberlites re-defined. Journal ofGeology 32, 223–228.

Dalton, J.A., Presnall, D.C., 1998. The continuum of primary carbonatite-kimberlitic meltcompositions in equilibrium with lherzolite: data from the system CaO–MgO–Al2O3–SiO2–CO2 at 6 GPa. Journal of Petrology 39, 1953–1964.

Dasgupta, R., Hirschmann, M.M., 2007. Effect of variable carbonate composition on thesolidus of mantle peridotite. American Mineralogist 92, 370–379.

Digonnet, S., Goulet, N., Bourne, J., Stevenson, R., Archibald, D., 2000. Petrology of theAbloviak aillikite dykes, New Québec: evidence for a Cambrian diamondiferousalkaline province in northeastern North America. Canadian Journal Earth Sciences 37,517–533.

Edwards, D., Rock, N.M.S., Taylor, W.R., Griffin, B.J., Ramsay, R.R., 1992. Mineralogy andpetrology of the Aries diamondiferous kimberlite pipe, central Kimberley Block,Western Australia. Journal of Petrology 33, 1157–1191.

Finnerty, A.A., Boyd, F.R., 1987. Thermobarometry for garnet peridotites: basis for thedetermination of thermal and compositional structure of the upper mantle. In:Nixon, P.H. (Ed.), Mantle Xenoliths. John Wiley and Sons, London, pp. 381–402.

Foley, S., 1992. Petrological characterization of the source components of potassicmagmas: geochemical and experimental constraints. Lithos 28, 187–204.

Francis, D., 2003. Cratonicmantle roots, remnants of amore chondritic Archeanmantle?Lithos 71, 135–153.

Fraser, K.J., Hawkesworth, C.J., 1992. The petrogenesis of group 2 ultrapotassickimberlites from Finsch Mine, South Africa. Lithos 28, 327–345.

Graham, I., Burgess, D., Bryan, D., Ravenscroft, P.J., Thomas, E., Dolye, B.J., Hopkins, R.,Armstrong, K.A., 1999. Exploration history and geology of the Diavik Kimberlites,Lac de Gras, Northwest Territories, Canada. In: Gurney, J. (Ed.), Proceedings of the7th International Kimberlite Conference. Red Roof Design, Cape Town, pp. 262–279.

Gudfinnsson, G., Presnall, D.C., 2005. Continuous gradations among primary carbona-titic, kimberlitic, melilititic, basaltic, picritic, and komatiitic melts in equilibriumwith garnet lherzolite at 3–8 GPa. Journal Petrology 46, 1645–1659.

Hirschmann, M.M., 2000. Mantle solidus: experimental constraints and the effects ofperidotite composition. Geochemistry, Geophysics, Geosystems 1.

Jaques, A.L., Milligan, P.R., 2004. Patterns and controls on the distribution ofdiamondiferous intrusions in Australia. Lithos 77, 783–802.

Jaques, A.L., Lewis, J.D., Smith, C.B., Gregory, G.P., Ferguson, J., Chapell, B.W., McCulloch,M.T., 1984. The diamond-bearing ultrapotassic (lamproitic) rocks of the WestKimberley region, western Australia. Proceedings of the Third InternationalKimberlite Conference 1, 225–254.

Kamenetsky, V.S., Kamenetsky, M.B., Sharygin, V.V., Faure, K., Golovin, A.V., 2007.Chloride and carbonate immiscible liquids at the closure of the kimberlite magmaevolution (Udachnaya-East kimberlite, Siberia). Chemical Geology 237, 384–400.

Kaminsky, F.V., Sablukov, S.M., Sablukova, L.I., Shchukin, V.S., Canil, D., 2002. Kimberlitesof the Wawa area. Canadian Journal Earth Sciences 39, 1819–1838.

Katz, R.F., Spiegelman, M., Langmuir, C.H., 2003. A new parameterization of hydrousmantle melting. Geochemistry, Geophysics, Geosystems 4, 1073.

Kirkley, M.B., Smith, H.S., Gurney, J.J., 1989. Kimberlite carbonates: a carbon and oxygenstable isotope study. In: Ross, J. (Ed.), Kimberlites and Related Rocks. GeologicalSociety of Australia, Sidney, pp. 264–281.

Larsen, L.M., Rex, D.C., 1992. A review of 2500 Ma span of alkaline-ultramafic, potassicand carbonatitic magmatism in West Greenland. Lithos 28, 367–402.

Le Roex, A.P., Bell, D.R., Davis, P., 2003. Petrogenesis of Group-I kimberlites fromKimberley, South Africa: evidence from bulk rock geochemistry. Journal ofPetrology 44, 2261–2286.

Lewis, T.J., Hyndman, R.D., Fluck, P., 2003. Heat flow, heat generation, and crustaltemperatures in the northern Canadian Cordillera: thermal controls of tectonics.Journal of Geophysical Research 108, 2316.

MacGregor, I.D., 1974. The system MgO–Al2O3–SiO2: solubility of Al2O3 in enstatite forspinel and garnet peridotite compositions. American Mineralogist 59, 110–119.

Malkovets, V.G., Griffin,W.L.,O'Reilly, S.Y.,Wood, B.J., 2007.Diamond, subcalcic garnet, andmantle metasomatism: kimberlite sampling patterns define the link. Geology 35,339–342.

McCammon, C., Kopylova, M.G., 2004. A redox profile of the slave mantle and oxygenfugacity control in the cratonic mantle. Contributions Mineralogy Petrology 148,55–68.

Mitchell, R.H., 1986. Kimberlites: mineralogy, geochemistry, and Petrology. PlenumPress, New York.

Mitchell, R.H., 1995. Kimberlites, orangeite, and related rocks. Plenum Press, New York.Mitchell, R.H., 2004. Experimental studies at 5–12 GPa of the Ondermatjie hypabyssal

kimberlite. Lithos 76, 551–564.Mitchell, R.H., 2008. Petrology of hypabyssal kimberlites: relevance to primary magma

compositions. Journal of Volcanology and Geothermal Research 174, 1–8.Mitchell, R.H., Bergman, S.C., 1991a. Petrology of Lamproites. Plenum Press, p. 447.Mitchell, R.H., Bergman, S.C., 1991b. Petrology of Lamproites. Springer, p. 440.Moore, K.R., Wood, B.J., 1998. The transition from carbonatite to silicate melts in the

CaO–MgO–SiO2–CO2 system. Journal of Petrology 39, 1943–1951.Nimis, P., Taylor, W.R., 2000. Single clinopyroxene thermometry for garnet peridotites.

Part 1. Calibration and testing of a Cr-in-Cpx barometer and an enstatite-in-Cpxthermometer. Contributions Mineralogy Petrology 139, 541–554.

Nowicki, T., Crawford, B., Dyck, D., Carlson, J., McElroy, R., Oshust, P., Helmstaedt, H.,2004. The geology of kimberlite pipes of the Ekati property, Northwest Territories,Canada. Lithos 76, 1–27.

O'Neill, H.S.C., Palme, H., 1998. Composition of the silicate Earth: implications foraccretion and core formation. In: Jackson, I. (Ed.), The Earth's Mantle composition,structure, and evolution. Cambridge University Press, Cambridge, pp. 3–126.

Palme, H., O'Neil, H.S.C. (2003) Cosmochemical estimates of mantle composition. In:Holland, H.D., and Turekian, K.K. (ed) The Mantle and Core, vol. 2. Elsevier-Pergamon, Oxford, pp 1-38.

Price, S.E., Russell, J.K., Kopylova, M.G., 2000. Primitivemagma from the Jericho Pipe, N.W.T.,Canada: constraints on primary kimberlite melt chemistry. Journal of Petrology 41,789–808.

Queen, M., Haeman, L.M., Hanes, J.A., Archibald, D.A., Farrar, R., 1996. 40Ar/39Arphlogopite and U–Pb perovskite dating of lamprophyre dykes from the eastern LakeSuperior region: evidence for a 1.14 Ga magmatic precursor to Midcontinent Riftvolcanism. Canadian Journal Earth Sciences 33, 958–965.

Rock, N.M.S., 1986. The nature and origin of ultramafic lamprophyres: alnoites andallied rocks. Journal of Petrology 27, 155–196.

Roeder, P.L., Emslie, R.F., 1970. Olivine–liquid equilibrium. Contributions MineralogyPetrology 29, 275–289.

Russell, J.K., Nicholls, J., 1988. Analysis of petrologic hypotheses with Pearce elementratios. Contributions to Mineralogy and Petrology 99, 25–35.

Schmidberger, S.S., Francis, D., 1999. Nature of the mantle roots beneath the NorthAmerican Craton: mantle xenolith evidence from Somerset Island kimberlites.Lithos 48, 195–216.

Sheppard, S.M.F., Dawson, J.B., 1975. Hydrogen, carbon, and oxygen isotope studies ofmegacryst and matrix minerals from Lesothan and South African kimberlites.Physics and Chemistry of the Earth 9, 747–763.

Shi, L., Francis, D., Ludden, J., Frederiksen, A., Bostock, M., 1998. Xenolith evidence forlithospheric melting above anomalously hot mantle under the northern CanadianCordillera. Contributions Mineralogy Petrology 131, 39–53.

Skinner, E.M.W., Viljoen, K.S., Clarke, T.C., Smith, C.B., 1994. The petrography, tectonicsetting and emplacement ages of kimberlites in the southwestern border region ofthe Kaapvaal craton, Prieska area. In: Meyer, H.O.A., Leonardos, O.H. (Eds.),Kimberlites, related rocks and mantle xenoliths. Special Publication, vol. IA.CPRM, pp. 80–97.

Smith, C.B., 1983. Pb, Sr, Nd isotopic evidence for the sources African Cretaceous kimberlites.Nature 304, 51–54.

Stripp, G.R., Field, M., Schumacher, J.C., Sparks, R.S.J., Cressey, G., 2006. Post-emplacementserpentinization and related hydrothermal metamorphism in a kimberlite fromVenetia, South Africa. Journal Metamorphic Petrology 24, 515–534.

Tappe, S., Jenner, G.A., Foley, S.F., Heaman, L., Besserer, B.A., Kjarsgaard, B.A., Ryan, B., 2004.Torngat ultramafic lamprophyres and their relation to the North Atlantic AlkalineProvince. Lithos 76, 491–518.

Tappe, S., Foley, S.F., Jenner, G.A., Kjarsgaard, B.A., 2005. Integrating ultramafic lamprophyresinto the IUGS classification of igneous rocks: rationale and implications. JournalPetrology 46, 1893–1900.

Tappe, S., Foley, S.F., Jenner, G.A., Heaman, L.M., Kjarsgaard, B.A., Romer, R.L., Stracke, A.,Joyce, N., Hoeffs, J., 2006. Genesis of ultramafic lamprophyres and carbonatites atAillik Bay, Labrador: a consequence of incipient lithospheric thinning beneath theNorth Atlantic craton. Journal of Petrology 47, 1261–1315.

Taylor, W.R., Green, D.H., 1988. Measurements of reduced peridotite–C–O–H solidus andimplications for redox melting of the mantle. Nature 332, 349–352.

Taylor, W.R., Tompkins, L.A., Haggerty, S.E., 1994. Comparative geochemistry of WestAfrican kimberlites: evidence for a micaceous kimberlite endmember of sub-lithospheric origin. Geochimica et Cosmochimica Acta 58 (19), 4017–4037.

Vasilenko, V.B., Zinchuk, N.N., Krasavchikov, V.O., Kuznetsova, L.G., Khlestov, V.V.,Volkova, N.I., 2002. Diamond potential estimation based on kimberlite majorelement chemistry. Journal of Geochemical Exploration 79 (2), 93–112.

Wilson, M.R., Kjarsgaard, B.A., Taylor, B., 2007. Stable isotope composition of magmaticand deuteric carbonate phases in hypabyssal kimberlite, Lac de Gras field,Northwest Territories, Canada. Chemical Geology 242, 435–454.

Woodland, A.B., Koch, M., 2003. Variation in oxygen fugacity with depth in the uppermantle. Earth and Planetary Science Letters 214, 295–310.

Yoder, H.S., 1986. Potassium-rich rocks: phase analysis and heteromorphic reactions.Journal of Petrology 27, 1215–1228.